Willy Cheung: Extending Agreement Detection: Dealing with Neutral and Uncertain Items
Detecting agreement (given two interlocuters where A makes a statement and B responds, determine if the responder B agrees/disagrees with A) is a relatively new NLP task involving naturally occurring data, and presents an interesting testbed for exploring pragmatic phenomenon. Working within the frame of a classification problem, prior work has seen relative success in binary classification. However, prior work fails to take into account some of the data, namely neutral and uncertain items. In my QP work, I explore ways to handle these types of items, and also present ideas for current feature engineering work.
Clinton Awai: Modalization of 'fall' in Hindi
In Hindi, the verb paṛ- ‘fall’ can be used as both a lexical verb, as in (1), as a light verb in a complex predicate, as in (2), and as a modal, as in (3).
(1) ghar paṛ-ā
‘The house fell’
(2) gārī cal paṛ-ī
train.F.NOM move fall-PERF.F.SG
‘the train began to move’ (McGregor 1933:112)
(3) yāsīn=ko ye kar-nā paṛ-ā
Yasin.M.SG=DAT this.SG.NOM do-INF.M.SG fall-PERF.M.SG
‘Yasin was obliged to do this’ (Bhatt et al. 2011:3)
One of the implications of the modal in sentence (3) is that the subject lacks agency and is under compulsion to perform the action.
The phenomenon of the verb ‘fall’ becoming a modal is not found outside of the Indo-Aryan language family, but the lack of agency, which can be seen as a first step towards the modality of obligation with paṛ-, does have parallels. Thus, the methodology of the traditional exploration of semantic change (as in Bréal (1900), Ullmann (1953), Buck (2008) and basically all textbooks that cover historical linguistics), involving seeking cross-linguistic parallels, can be useful here.
There are parallels to this non-volitional feature of ‘fall’ that is found elsewhere. In some languages, like Korean, ‘fall’ has been recruited to take on the function of a passive auxiliary.
I report here on the findings of this investigation. Besides offering some insight into this semantic shift within the history of Hindi, this research has uncovered an as-yet undescribed pathway — not to be found in Heine & Kuteva 2004, for instance – from a lexical verb to a modal of obligation.
Daniel Puthawala: Stripping in HTLCG
In this project, I examine the elliptical phenomenon of stripping in English. In contrast to other accounts of stripping, I take a categorial grammatical approach to stripping, and draw upon an analysis of Gapping in HTLCG by Kubota & Levine (2016). Though the Kubota and Levine (2016) account was custom-built to handle gapping, I will show how it can be adapted, with minimal additions, to straightforwardly obtain both the straightforward semantics of simple stripping cases, as well as the complex scopal relations of negation and auxiliaries in more complicated cases.
The success of my account of stripping also has theoretical and empirical ramifications beyond the narrow boundaries of stripping. First, my success in this project gives evidence that the mechanics that underly Kubota and Levine's (2016) analysis of gapping are more generally useful than just in cases of gapping. Secondly, my analysis raises questions as to how formally and distributionally similar stripping is to coordination, gapping, and sluicing.
There will be a reception following the talks.